Hindawi Publishing Corporation BioMed Research International Volume 2013, Article ID 924137, 15 pages http://dx.doi.org/10.1155/2013/924137 Research Article Mixing Energy Models in Genetic Algorithms for On-Lattice Protein Structure Prediction Mahmood A. Rashid, 1,2 M. A. Hakim Newton, 1 Md. Tamjidul Hoque, 3 and Abdul Sattar 1,2 1 Institute for Integrated & Intelligent Systems, Science 2 (N34) 1.45, 170 Kessels Road, Nathan, QLD 4111, Australia 2 Queensland Research Lab, National ICT Australia, Level 8, Y Block, 2 George Street, Brisbane, QLD 4000, Australia 3 Computer Science, 2000 Lakeshore drive, Math 308, New Orleans, LA 70148, USA Correspondence should be addressed to Mahmood A. Rashid; mahmood.rashid@gmail.com Received 30 April 2013; Revised 16 August 2013; Accepted 19 August 2013 Academic Editor: Tatsuya Akutsu Copyright © 2013 Mahmood A. Rashid et al. his is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Protein structure prediction (PSP) is computationally a very challenging problem. he challenge largely comes from the fact that the energy function that needs to be minimised in order to obtain the native structure of a given protein is not clearly known. A high resolution 20×20 energy model could better capture the behaviour of the actual energy function than a low resolution energy model such as hydrophobic polar. However, the ine grained details of the high resolution interaction energy matrix are oten not very informative for guiding the search. In contrast, a low resolution energy model could efectively bias the search towards certain promising directions. In this paper, we develop a genetic algorithm that mainly uses a high resolution energy model for protein structure evaluation but uses a low resolution HP energy model in focussing the search towards exploring structures that have hydrophobic cores. We experimentally show that this mixing of energy models leads to signiicant lower energy structures compared to the state-of-the-art results. 1. Introduction Proteins are essentially sequences of amino acids. hey adopt speciic folded three-dimensional structures to perform spe- ciic tasks. However, misfolded proteins cause many critical diseases such as Alzheimer’s disease, Parkinson’s disease, and cancer [1, 2]. Protein structures are important in drug design and biotechnology. Protein structure prediction (PSP) is computationally a very hard problem [3]. Given a protein’s amino acid sequence, the problem is to ind a three-dimensional structure of the protein such that the total interaction energy amongst the amino acids in the sequence is minimised. he protein folding process that leads to such structures involves very complex molecular dynamics [4] and unknown energy fac- tors. To deal with the complexity of PSP in a hierarchical way, researchers have used discretised lattice-based structures and simpliied energy models [5–7]. here are a large number of existing search algorithms that attempt to solve the PSP problem by exploring feasi- ble structures called conformations. For the low resolution hydrophobic-polar (HP) energy model, a memory based local search algorithm [8, 9], a population-based genetic algorithm [10], and a hydrophobic core directed local search method [11] reportedly produced the state-of-the-art results on the face-centred-cubic (FCC) lattice. For the high resolu- tion Berrera 20 × 20 energy matrix (henceforth referred to as BM energy model) [12–14] produces the state-of-the-art results. Nevertheless, the challenges in PSP largely remain in the fact that the energy function that needs to be minimised in order to obtain the native structure of a given protein is not clearly known. A high resolution 20 × 20 energy model (such as BM) could better capture the behaviour of the actual energy function than a low resolution energy model (such as HP). However, the ine grained details of the high resolution interaction energy matrix are oten not very informative for guiding the search. Pairwise contributions that have large magnitudes could be overshadowed by the accumulation of pair-wise contributions having small magnitudes or opposite signs. In contrast, a low resolution energy model could efec- tively bias the search towards certain promising directions